E(y)1/TAF9 mediates the transcriptional output of Notch signaling in Drosophila

e(y)1 is required for Notch-signaling-mediated follicle cell transition from mitotic cycle to endocycle during Drosophila oogenesis

To identify genes that modulate Notch signaling, we performed a large-scale in vivo RNAi screen for defects in the expression of Notch target genes in Drosophila melanogaster follicle cells. The follicle cell epithelium shows a temporal Notch activation pattern during stages 7–10A of oogenesis, with both negative and positive targets identified (Sun and Deng, 2007; Sun and Deng, 2005). We tested around two thousand Transgenic RNAi Project (TRiP) RNAi lines (Ni et al., 2009; Ni et al., 2008; Ni et al., 2011), which contain either long double-strand (ds) hairpin (Ni et al., 2009; Ni et al., 2008) or short hairpin RNAs (Ni et al., 2011) (both referred to as RNAi) under the control of the upstream activating site (UAS). These lines were crossed to the flip-out Gal4 driver (Ito et al., 1997; Pignoni and Zipursky, 1997) to generate random follicle-cell RNAi clones that were marked by the expression of green and red fluorescent proteins (GFP and RFP, respectively). The follicle-cell RNAi clones were then screened for defects in the expression pattern of the negative Notch target Cut, a homeodomain-containing transcription factor (Sun and Deng, 2005). Cut is normally expressed in early oogenesis (stages 1–6) and then downregulated at stage 7 upon Notch activation in follicle cells (Sun and Deng, 2005). Failure to downregulate Cut at stage 7 suggests a defect in Notch signaling. From this screen, we found that knockdown of the gene CG6474 [enhancer of yellow 1 (e(y)1)], which encodes a homolog of TAF9, a core component of the TFIID transcription initiation complex (Soldatov et al., 1999; Thomas and Chiang, 2006), resulted in prolonged Cut expression in 74% (n = 72) of follicle cell RNAi clones at stages 7–8 (Fig. 1A), suggesting a Notch defect in e(y)1-depleted follicle cells.

To confirm that E(y)1 is required for Notch signaling in follicle cells, we examined the expression of zinc-finger protein Hnt, a positive follicle-cell-specific Notch target, which is normally expressed in the entire follicular epithelium from stages 7–10A (Sun and Deng, 2007). Downregulation of Hnt in mid-oogenesis (stages 7–10A) is an indication of disrupted Notch signaling in follicle cells (Sun and Deng, 2007). Indeed, Hnt downregulation was detected in 68% (n = 51) of e(y)1-knockdown follicle cells (Fig. 1B). Both Cut upregulation and Hnt downregulation were detected in an independent e(y)1-RNAi line (no. 6474R-1) from the National Institute of Genetics (NIG) that was used to knock down e(y)1 expression in mid-staged egg chambers (supplementary material Fig. S1A; data not shown).

To further exclude potential off-target effects on Notch signaling caused by e(y)1-RNAi, we set out to analyze the phenotypes of e(y)1 mutants. An allele named e(y)1¹⁹⁰ was generated through imprecise excision of a P-element that had been inserted into the 5′-UTR of the e(y)1 locus (Fig. 2A; see Materials and Methods for details). This allele was both hemizygous and homozygous lethal during early larval stages and could be rescued to adulthood by a duplication [Dp(1;3)DC335 or Dp(1;3)DC336] harboring the e(y)1 genomic sequence (data not shown). Flippase-flippase recognition target (FLP-FRT)-induced follicle cell mitotic clones of e(y)1¹⁹⁰ showed upregulated levels of Cut and downregulated levels of Hnt during stages 7–8 when compared with neighboring wild-type cells (Fig. 2B–C″), supporting our findings in e(y)1 knockdown cells.

Notch-induced transcription activation can be assessed in a more direct manner by using transgenic reporters. Both E(spl)mβ-CD2 (de Celis et al., 1998; Furriols and Bray, 2001) and E(spl)m7-lacZ (Assa-Kunik et al., 2007; Pines et al., 2010), which contain the Su(H) binding sites, are expressed in follicle cells during mid-oogenesis upon Notch activation. Indeed, the expression of these two reporters was substantially reduced in e(y)1-RNAi follicle cells (Fig. 1C,D). In addition, the expression of another follicle-cell-specific Notch reporter broad-early-enhancer-lacZ (brE-lacZ), recently identified by our laboratory (Jia et al., 2014), was also decreased in e(y)1-depleted follicle cells (supplementary material Fig. S1B). These results suggest that e(y)1 is required for proper transcriptional regulation of the Notch target genes in follicle cells.

A developmental consequence of Notch activation in a follicle cell during stage 7 is the induction of a switch from the mitotic divisions of early oogenesis to endoreplication cycles (endocycles) (the M/E transition) during mid-oogenesis (Deng et al., 2001; López-Schier and St Johnston, 2001). To determine whether e(y)1 is required for this Notch-dependent event, we examined nuclei size and expression of mitotic markers in mosaic egg chambers containing follicle-cell clones that expressed RNAi against e(y)1. Knockdown of e(y)1 resulted in smaller and more densely distributed nuclei than those of the neighboring wild-type follicle cells, suggesting a failure in the M/E switch (supplementary material Fig. S2A′). To determine whether e(y)1-knockdown follicle cells remained in the mitotic cycle, we monitored two mitotic markers – phosphorylated histone H3 and Cyclin B (CycB) – by staining the mosaic egg chambers. In wild-type egg chambers, phosphorylated histone H3 and CycB oscillate in early mitotic cycles up to stage 6 and are absent during endoreplication stages (stages 7–10A) (Deng et al., 2001; Shcherbata et al., 2004), whereas in e(y)1-knockdown follicle cells, both markers showed random expression after stage 6 (supplementary material Fig. S2B,C). Taken together, these data indicate that e(y)1 is potentially required for the Notch-dependent M/E switch in Drosophila follicle cells.

Disruption of e(y)1 affects the expression of Notch target genes in the wing imaginal disc and cultured S2 cells

To assess whether e(y)1 also modulates Notch signaling in other tissues, we examined the expression of two well-documented Notch target genes, wg and cut, in the wing imaginal disc. During wing disc development, the Notch signaling pathway is activated at the dorsal–ventral boundary by an interaction between the Notch receptor on one cell and its ligand (Delta or Serrate) on the neighboring cell (Diaz-Benjumea and Cohen, 1995; Doherty et al., 1996). Because e(y)1 mutant cells in the wing disc generated by FLP-FRT mosaic analysis displayed very poor viability (supplementary material Fig. S3B,B′), we employed the RNAi line to study the role of e(y)1 in Notch signaling during wing development. To circumvent early larval lethality of engrailed-Gal4 (en-Gal4)-induced e(y)1-RNAi (data not shown), we used the temporally controlled Gal4-Gal80ts system (McGuire et al., 2003) to express UAS-e(y)1-RNAi in the posterior compartment of a wing disc marked by GFP expression. The control discs did not show detectable changes in Wg (Fig. 3I) or Cut expression (Fig. 3J) (compare the GFP-positive posterior region with the GFP-negative anterior region). However, in en-Gal4, UAS-GFP; tub-Gal80ts>e(y)1-RNAi discs, levels of Wg or Cut in the GFP-expressing posterior compartment were decreased as compared with those in the anterior compartment (Fig. 3K,L). These results demonstrate that e(y)1 is involved in the expression of Notch target genes in the wing disc.

To determine whether E(y)1 is necessary for the transcription of Notch target genes when Notch signaling is artificially induced, we employed a well-documented transient Notch activation system in cultured S2 cells by transfecting a Notch-expressing plasmid, pMT-Notch (Krejcí and Bray, 2007; Yu et al., 2013). In this assay, the overexpressed full-length Notch can be cleaved to produce a constitutively active form of NICD (denoted by N^ICD) upon treatment with EDTA. N^ICD translocates into the nucleus and physically interacts with Su(H), which activates the transcription of Notch target genes. In this transient Notch activation system, assuming that E(y)1 is required for regulating Notch target gene expression, the induced expression of Notch target genes by pMT-Notch should be compromised by a perturbation of e(y)1. As expected, the introduction of interfering dsRNAs against e(y)1 [dse(y)1], but not against GFP (dsGFP), into the Notch-activating S2 cells led to a significant reduction in expression of the primary endogenous Notch target genes E(spl)m7 and E(spl)m8 (also known as E(spl)m7-HLH and E(spl)m8-HLH, respectively) (Fig. 5D,E). These results suggest that E(y)1 is also required for artificially induced Notch signaling in cultured S2 cells.

e(y)1 genetically interacts with components of the Notch pathway

To determine whether e(y)1 is functionally involved in Notch signaling in the wing imaginal disc, we knocked down e(y)1 through use of a tissue-specific Gal4 driver, C96-Gal4, which is expressed in the cells of the dorsal–ventral boundary of the wing disc, the precursor of the adult wing margin (Gustafson and Boulianne, 1996; Saj et al., 2010). Defective Notch signaling in the dorsal–ventral boundary leads to loss of wing margins (de Celis et al., 1996). Depletion of e(y)1 under the C96-Gal4 driver also led to partial wing margin loss in 96.9% (n = 64) of wings (Fig. 3B), similar to the Notch loss-of-function phenotype. This notched wing margin phenotype was also observed in mosaic adult flies in which e(y)1¹⁹⁰ random clones were induced at larval stages by heat-shock-induced Flp (hsFLP) (Fig. 2D). Notch heterozygous animals (N^264-39/+) showed a mild wing margin notching in the distal region in 48.5% (n = 101) of wings (Fig. 3C). Combination of C96-Gal4>e(y)1-RNAi and N^264-39/+ resulted in a synergistically enhanced notched wing phenotype in terms of both penetrance and expressivity (100%, n = 32; Fig. 3D). Furthermore, although e(y)1 heterozygous flies [e(y)1^190/+] did not show any visible wing defects (data not shown), incorporation of a copy of the e(y)1¹⁹⁰ mutation into Notch heterozygotes greatly enhanced the wing margin notching phenotype in both expressivity (Fig. 3F compared with Fig. 3E) and penetrance [increasing from 48.5% (n = 132; Fig. 3E) to 90% (n = 60; Fig. 3F)]. Similarly, e(y)1¹⁹⁰ also displayed a synergistic interaction with another Notch allele, N¹ (Fig. 3G,H). In addition to dorsal–ventral boundary formation in the wing disc, Notch signaling also plays an important role in cell fate determination through lateral inhibition (Artavanis-Tsakonas and Muskavitch, 2010). Indeed, we observed lateral specification defects in bristle and wing vein development when e(y)1 was depleted (supplementary material Fig. S3C–H). These genetic experiments suggest that e(y)1 functionally interacts with the Notch pathway during Drosophila wing development.

E(y)1 mediates the transcriptional output of Notch signaling through association with NICD and Su(H)

The Notch signaling pathway is delicately regulated at multiple levels, including ligand and receptor post-translational modification, endocytosis and vesicle trafficking, and epigenetic regulation of target gene expression (Fischer et al., 2006; Fortini and Bilder, 2009; Tien et al., 2009; Yu et al., 2013; Zeng et al., 2013). Interaction between Notch ligand and receptor induces two consecutive proteolytic cleavages of the Notch receptor, S2 and S3, which results in the release of NICD (Guruharsha et al., 2012). Translocation of the cleaved NICD to the nucleus activates transcription of the target genes through formation of a ternary complex with the DNA-binding protein Su(H) and the coactivator Mam (Guruharsha et al., 2012). e(y)1 putatively encodes a core component of the TFIID transcription initiation complex, which is required for polymerase-II-mediated gene expression (Dynlacht et al., 1991). Thus, two possibilities for its role in Notch signaling exist – first, E(y)1 is required for the transcriptional expression of a core component or a regulator of the Notch pathway; second, E(y)1 is directly engaged in the transcriptional control of Notch target genes that are mediated by the core NICD–Su(H)–Mam complex. To distinguish between these two possibilities, we conducted epistatic analyses in follicle cells by expressing transgenes that encode components acting in different steps of the Notch pathway. These transgenic constructs include a membrane-tethered active form of Notch, N^EXT; a constitutively active form of Notch, N^ICD (Rebay et al., 1993); and a fusion protein Su(H)–VP16 comprising the DNA-binding domain of Su(H) and the strong activation domain of viral transcription-factor protein VP16 (Kidd et al., 1998; Shyu et al., 2009). N^EXT is produced after S2 cleavage and N^ICD after S3 cleavage. Because the majority of Notch regulators and core components function upstream of either S2 cleavage or S3 cleavage, if E(y)1 is required for the expression of one of the necessary proteins, expression of N^EXT and/or N^ICD would rescue the defects caused by e(y)1 depletion. However, if E(y)1 is directly involved in the expression of Notch target genes, N^EXT and/or N^ICD would have no effect, but Su(H)–VP16 might activate Notch target gene expression in the absence of e(y)1, because Su(H)–VP16 is a more potent transcription activator than Su(H) and can activate a Notch target independently of Notch (Kidd et al., 1998). Because, in addition, the VP16 activation domain can directly interact with multiple components of different general transcription factor complexes (Hall and Struhl, 2002), we speculate that expression of certain Notch targets by Su(H)–VP16 might bypass E(y)1. Expression of these constructs can therefore distinguish whether the involvement of E(y)1 is at the level of target gene expression or component expression. Cut upregulation in e(y)1-knockdown follicle cells was employed as a readout in this epistatic analysis. During stages 7–8 of oogenesis, 76% (n = 66) of e(y)1-RNAi follicle cell clones showed Cut upregulation (Fig. 4A). Similar percentages of Cut upregulation were detected when N^EXT (79%, n = 46; Fig. 4B) or N^ICD (70%, n = 82; Fig. 4C) was co-expressed with e(y)1-RNAi. By contrast, Cut upregulation resulting from e(y)1-RNAi was dramatically reduced by co-expression of Su(H)–VP16 (24%, n = 57; Fig. 4D). These results suggest that the involvement of E(y)1 in Notch signaling is probably not related to the expression of Notch regulators or core components upstream of the release of NICD. We employed similar epistatic analysis for the function of E(y)1 in the adult wing. As seen in Fig. 3B, C96-Gal4>e(y)1-RNAi adults had notched wing margins (Fig. 4F). Although co-expression of N^ICD did not alleviate this defect (Fig. 4G), C96-Gal4>e(y)1-RNAi animals expressing Su(H)–VP16 showed greatly alleviated wing margin defects (Fig. 4H). C96-Gal4>Su(H)–VP16 alone displayed typical Notch gain-of-function phenotypes during Drosophila wing development (supplementary material Fig. S4). Taken together, these epistatic results in both follicle cells and developing wings suggest that E(y)1 acts downstream of NICD and is engaged in transcriptional activation of Notch target genes.

Previous studies have shown that specific transcription factors can directly interact with specific TAFs to recruit the transcriptional machinery to their target genes (Kalogeropoulou et al., 2010; Wright and Tjian, 2009). Su(H) is the executive core transcription factor of Notch signaling. Genetic analyses encouraged us to hypothesize that Su(H) physically interacts with E(y)1 to recruit the transcriptional machinery to initiate Notch target gene expression. To test this hypothesis, we turned to the Drosophila S2 cell system again. We co-transfected two constructs to express Flag-tagged E(y)1 and Myc-tagged Su(H), respectively, both under the control of the Actin5C promoter in S2 cells. As expected, Flag-tagged E(y)1 was able to immunoprecipitate Myc-tagged Su(H) (Fig. 5A). The NICD–Su(H)–Mam ternary complex has been reported to regulate Notch target gene activation in many organisms, and we therefore asked whether E(y)1 is also physically associated with NICD to promote Notch activation. Indeed, the association between E(y)1 and NICD was detected in a co-immunoprecipitation assay (Fig. 5B). Thus, E(y)1 mediates the transcriptional output of Notch signaling by interacting with NICD and Su(H) biochemically.

Requirement of other TAFs in Notch signaling

The results described thus far indicate that E(y)1 is required for the NICD–Su(H)–Mam-complex-mediated transcriptional program of Notch signaling. E(y)1, the Drosophila homolog of human TAF9, was initially identified as one of the core components of the TFIID complex (Dynlacht et al., 1991; Goodrich et al., 1993). To determine whether E(y)1 is a specific factor of the TFIID complex for Notch signaling and/or whether all components of the TFIID complex are required for Notch target gene activation, we examined Notch activity during Drosophila wing development after depleting other components of the TFIID complex through RNAi obtained from the TRiP. Knockdown of TAF12 by the en-Gal4 driver did not obviously affect Notch activity, as revealed by the expression of Wg (Fig. 6A), and consistently, the adult wings of C96- Gal4>TAF12-RNAi animals had normal wing margins (Fig. 6D). However, ablation of TAF1 under the en-Gal4 driver visibly decreased the level of Wg in the posterior compartment of the wing disc (Fig. 6B), and expectedly, overexpression of TAF1-RNAi induced by C96-Gal4 caused notched wing margins in adult flies (Fig. 6E). Furthermore, we tested several RNAi lines corresponding to specific components of the TFIID complex (TAF2, TAF5 and TAF6) available at the Vienna Drosophila RNAi Center under the same C96-Gal4 driver, and all of them showed notched wing margin defects with different expressivity (data not shown). These results suggest that all components of the TFIID complex we tested, except TAF12, are required for Notch signaling, at least during Drosophila wing development. The e(y)1, e(y)2 and e(y)3 genes were genetically identified as the respective mutations enhance the phenotype of the y² mutation (Georgiev and Gerasimova, 1989; Georgiev et al., 1990), but only E(y)1 is believed to be a component of the TFIID complex. To investigate whether e(y)2 or e(y)3 plays a similar role in Notch signaling as that of e(y)1, we compromised e(y)2 function by expression of e(y)2-RNAi and found that it was not required for Notch activity during wing development (Fig. 6C,F), implying that unlike e(y)2, e(y)1 is specifically required for Notch signaling.

Generation and characterization of the e(y)1¹⁹⁰ allele

Fly P element w¹¹¹⁸ P{GT1}e(y)1^BG00948 was inserted 78 base pairs (bps) upstream of the start codon of the e(y)1 gene. Isogenized P-element flies that are both homozygous and hemizygous viable and fertile were used for generation of the e(y)1¹⁹⁰ allele by a standard P-element-mediated imprecise jump-out strategy. Excisions of P{GT1}e(y)1^BG00948 that were hemizygous lethal were sequenced by the use of a pair of primers (5′-CATAAGCTCACCGATTTC-3′ and 5′-CCTCCATCTTGAGATCTC-3′) that flanked the genomic region of the P{GT1}e(y)1^BG00948 insertion site. The e(y)1¹⁹⁰ mutation is a deletion of 339 bps including part of the 5′-UTR, exon 1, intron 1 and part of exon 2 of the e(y)1 locus (see Fig. 2A).

Fly stocks and genetics

For clonal analyses, the mutation of e(y)1¹⁹⁰ was recombined with hsFLP¹²² and FRT19A on the X chromosome. Two independent UAS-e(y)1-RNAi lines were used to knock down e(y)1 levels in this paper – e(y)1^{TRiP#HMS00336} from the Bloomington Drosophila Stock Center and e(y)1^6474R-1 from the National Institute of Genetics, Japan. To induce random clones in follicle cells, the flip-out Gal4 driver [hsFLP; act>CD2 (or y⁺)>Gal4] was crossed to flies carrying an RNAi or an overexpression construct under the control of the UAS promoter or combination of interest. Their adult progeny were heat-shocked twice at 37°C for 30 minutes and then cultured in a wet-yeast-pasted vial at 29°C for 2–3 days before dissection. w ubi-GFP M(1)^osp FRT19A was a gift from Richard S. Mann (Estella and Mann, 2010). Other fly stocks and genotypes related to the figures listed are listed in supplementary material Table S1.

Immunocytochemistry

Antibody staining was performed as previously described (Deng et al., 2001). Antibodies from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA) were against Cut (1∶20), CycB (1∶10), Wg (1∶20) and Hnt (1∶15). Other antibodies used were against phosphorylated histone H3 (rabbit, 1∶400; Upstate), CD2 (mouse, 1∶50; Serotec) and β-Galactosidase (rabbit, 1∶2000; Sigma). Nuclei were co-stained with DAPI (1∶1000, Invitrogen) for 15 minutes at room temperature. Images were captured on a Zeiss LSM-510 confocal microscope in the Biological Science Imaging Resource Facility at Florida State University. Figures were processed and arranged in Adobe Illustrator.

S2 cell culture, transfection and RNAi assays

Drosophila S2 cells were cultured at room temperature in Hyclone serum-free insect cell culture media (Roche). Transfection of S2 cells was performed using FuGENE HD transfection kit reagents (Roche) following the manufacturer's instructions. The constructs used for transfections are available upon request. The S2 cells were normally harvested 48 hours after transfection for further experiments. Full-length Notch expression was induced by using 500 µM CuSO₄ for 24 hours after transfection with pMT-Notch (Fehon et al., 1990). dsRNA was prepared with the RiboMAX large-scale RNA production system-T7 kit (Promega) as previously described (Huang et al., 2011). Primers used were – e(y)1 forward 5′-TTAATACGACTCACTATAGGGGAGATCATGTCCATCCTGAAGGAG-3′ and e(y)1 reverse 5′-TTAATACGACTCACTATAGGGGAGACGCTAGTTGGTCACAAACTC-3′; control GFP forward 5′-TTAATACGACTCACTATAGGGGAGAATGGTGAGCAAGGGCGAGGAGCTG-3′ and GFP reverse 5′-TTAATACGACTCACTATAGGGGAGACTTGTACAGCTCGTCCATGCCGAGAG-3′ (bases in bold indicate T7 promoter sequences). For a 6-well plate, cells in 2 ml of medium in each well were treated with 15 µg dsRNA for 4 days before plasmid transfection or induction with CuSO₄.

Co-immunoprecipitation assay

Total protein extracts from S2 cells were prepared in immunoprecipitation lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA pH 7.4, 1% Triton X-100, 0.1% SDS) in the presence of protease inhibitors [1 mM PMSF; protease inhibitor cocktail (Calbiochem)]. For co-immunoprecipitation assays, extracts were incubated with specific antibodies and protein A/G agarose beads (Abmart) at 4°C overnight before washes and elution. Immunoprecipitates were boiled in 2×SDS loading buffer for elution from the beads. Rabbit anti-Flag (1∶100, Sigma) and rabbit anti-Myc (1∶200, Sigma) antibodies were used for co-immunoprecipitation experiments; mouse anti-E(y)1 (1∶1000, Sigma), rabbit anti-Flag (1∶1000, Sigma), rabbit anti-Myc (1∶2000, Sigma) and mouse anti-Tubulin (1∶2000, Sigma) antibodies were used for western blots.